The Brain Starts Making Decisions Far Earlier Than Scientists Thought

The classical view of how the brain makes decisions is elegantly simple: sensory information flows in through dedicated early regions, climbs the neural hierarchy, and arrives at the prefrontal cortex, where a choice is made. Everything downstream of that top-level judgment simply executes. A new study turns that model on its head.

Researchers at the University of Illinois Urbana-Champaign have discovered that decision-making signals appear as early as the primary somatosensory cortex (S1) — the brain's very first cortical relay for touch and movement sensation. Published in the Proceedings of the National Academy of Sciences, the findings challenge one of the foundational assumptions behind modern artificial intelligence: that decisions are made at the top of a hierarchy after information flows all the way up from the senses.

The team, led by Professor Yurii Vlasov and doctoral student Alex Armstrong, recorded neural activity in mice navigating a virtual reality corridor using their whiskers. The experiment created what the paper calls an "information bottleneck" — a controlled chokepoint that forced all tactile decision-relevant input through a single, measurable channel in the brain. What they observed contradicted the standard picture entirely.

During the evidence-accumulation phase of each decision, neural activity in S1 did not passively relay sensory data upward. Instead, the high-dimensional spiking activity across hundreds of individual neurons collapsed into a single latent variable, followed by a slow, synchronized ramp across the entire cortical column. This ramp-up pattern is a signature previously associated with downstream decision areas, not with primary sensory cortex. More striking still, the mechanism turned out to be driven by top-down feedback: higher brain regions — premotor and frontal areas traditionally associated with deciding — were actively sending signals back into S1, shaping what it encoded during the critical decision window.

"Decision-making is not solely relying on unidirectional feed-forward processes as previously thought," the paper states. "S1 appeared to be dynamically modulated by top-down regulation, engaged by higher-level brain regions via feedback loops."

The finding carries significant implications for artificial intelligence. Most modern AI systems — particularly convolutional neural networks used in computer vision — process information in a strict feedforward cascade, where each layer passes its output upward without any systematic influence flowing back down. If the brain's efficiency depends on these rapid bidirectional feedback loops, current AI architectures may be missing a crucial computational principle. The study suggests that building machines that truly think like brains may require fundamentally rethinking how information flows through artificial neural networks.